Research map / Architectures and efficiency / RNNs and normalization
Multi-ReduNet: Interpretable Class-Wise Decomposition of ReduNet
Fengrong Li, Delin Chu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| class-wise decomposition (Multi-ReduNet) instead of ReduNet | ReduNet · Reuters SVM | accuracy: 0.802 -> 0.985 | Improved |
| class-wise decomposition instead of ReduNet | ReduNet · DrivFace and ARCENE | accuracy: 0.43-0.46 -> 0.73-1.00 | Improved |
| class-wise decomposition instead of ReduNet | ReduNet · MNIST SVM | accuracy: 0.906 -> 0.842 | Hurt |
| class-wise decomposition instead of ReduNet | ReduNet · Fashion-MNIST SVM | accuracy: 0.858 -> 0.801 | Hurt |
| add LastNorm (defer normalization to final layer) (ablation) | Multi-ReduNet · six datasets best accuracy | best accuracy: +0.2 pp average, +1.3 pp on Fashion-MNIST | Improved |
| add LastNorm to reduce hyperparameter sensitivity (ablation) | Multi-ReduNet · learning rate range | accuracy range: 9.0 pp -> 6.4 pp | Improved |
| class-wise decomposition for training speed | ReduNet · six datasets training time | training time: 1.4-2.6x speedup, about 2x average | Improved |
| Random Forest soft membership predictor (ablation) | ReduNet · accuracy and efficiency | accuracy | No clear effect |
| Multi-ReduNet-LN instead of ReduNet | ReduNet · Mice Protein SVM eta 0.5 | accuracy: 0.929 -> 0.605 | Hurt |
| Multi-ReduNet-LN instead of ReduNet | ReduNet · CIFAR-10 KNN eta 0.5 | accuracy: 0.182 -> 0.276 | Improved |
| Multi-ReduNet-LN instead of LDA | LDA · ARCENE | accuracy: 0.878 -> 0.829 | Hurt |
| Multi-ReduNet-LN instead of global PCA + SVM | Global PCA + SVM · Reuters | accuracy: 97.5% -> 98.8% | Improved |
Closest papers by meaning
- QPM: Discrete Optimization for Globally Interpretable Image Classification (ICLR 2025 · similarity 0.84)
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.84)
- Separation Power of Equivariant Neural Networks (ICLR 2025 · similarity 0.83)
- PICNN: A Pathway towards Interpretable Convolutional Neural Networks (AAAI 2024 · similarity 0.83)
- Unlearning-based Neural Interpretations (ICLR 2025 · similarity 0.83)
- Reducing Class-wise Confusion for Incremental Learning with Disentangled Manifolds (CVPR 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
